Explainable Failure Forecasting from Storage Device Telemetry under Dataset Shift

Abstract

Storage devices exhibit different failure precursors across firmware versions, manufacturers and production workloads. An evaluation that overlooks these constraints may reach conclusions that cannot be reproduced. The planned investigation seeks to test the stability and actionability of explainable failure forecasts when telemetry distributions shift. Relevant methodological and data-quality literature informs the design of the sampling and validation strategy [3–7]. The protocol specifies a time-split observational evaluation using anonymised device histories and independently confirmed service events. Comparators will include static health scores, survival-style baselines and calibrated explainable classifiers. Prior work on storage operations provides technical context for validation [1], while complementary systems evidence motivates service-continuity checks [2]. The design prioritises precise operational definitions and a straightforward baseline before introducing more complex alternatives. Statistical and computational sources help define the reproducibility checks and comparison procedures [8–12]. The primary outcomes are false alarms per device-month, early-warning lead time, calibration and plausibility of salient features. Potential biases include unreliable fault labels, censored histories, cohort leakage and explanation drift. Primary analyses would be reported by the main deployment or participant subgroup so that averaged results do not conceal systematic failures. Additional research informs the prespecified stress tests and limits on interpretation [13–17]. The intended contribution is an auditable research protocol rather than an unsupported claim of effectiveness; related work frames its transferability limits [18–23].